The aim of this study was to evaluate completeness of tuberculosis notification in Bichat Claude-Bernard University Hospital and to evaluate whether misclassification of atypical mycobacterial infection could have contributed to the inaccuracy of tuberculosis notification. Data from Microbiology Laboratory of the hospital and statutory notifications were compared. From 1 January 1994 to 31 December 1995, 299 tuberculosis cases were diagnosed in the Microbiology Laboratory and 316 cases were notified as tuberculosis. Notification rate for laboratory-documented tuberculosis was 57.5%, was significantly higher in cases with positive acid fast bacilli smear (75%) than without this feature (45%) and was similar in HIV-positive (59.4%) and HIV-negative (63.5%) patients. Among notified cases, diagnosis was established by laboratory proofs in only 54.4% and by clinical signs in 45.6%. Three cases with positive smear and culture growing atypical mycobacteria were wrongly notified. Notification of laboratory-documented tuberculosis was higher than that observed in a previous study in the same hospital, suggesting that the rise of tuberculosis incidence reported in our country could be partially artificial. Nevertheless, extent of notification remains insufficient and needs to be improved by combining microbiological data with current system of notification.